The growing cyber threat landscape necessitates robust defenses, with Network Intrusion Detection Systems (NIDS) at the forefront. Leveraging Deep Neural Networks (DNN) has significantly improved detection accuracy in NIDS. Nonetheless, the inherent data imbalance between malicious and normal network traffic impairs the efficacy of DNN-based methods. Traditional approaches employ Generative Adversarial Networks (GANs) to mitigate this by generating minority class samples. However, these methods often struggle with the scarcity of specific data types during training, leading to low-quality synthetic samples and sub-optimal detection performance. Addressing this, we introduce the Synchronous Classifier Wasserstein Generative Adversarial Network (SC-WGAN). This novel model extends the WGAN framework by integrating a classifier that processes all traffic data types. This classifier collaborates with the discriminator during the training process, guiding the generator to produce higher-quality synthetic samples. We evaluate SC-WGAN’s performance on three well-known open-source benchmark datasets. Additionally, we compare it against contemporary GAN-based solutions tackling data imbalance. Our findings reveal that SC-WGAN, surpasses existing methods in generating more representative samples and enhancing NIDS detection accuracy.